2 <!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"
3 "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd"
8 <article id="faq" xreflabel="Frequently Asked Questions">
9 <?dbhtml filename="faq.html"?>
12 <title>Frequently Asked Questions</title>
18 <ulink url="http://www.fsf.org">FSF</ulink>
23 <!-- FAQ starts here -->
26 <!-- General Information -->
27 <qandadiv id="faq.info" xreflabel="General Information">
28 <title>General Information</title>
30 <qandaentry id="faq.what">
31 <question id="faq.what.q">
36 <answer id="faq.what.a">
38 The GNU Standard C++ Library v3 is an ongoing project to
39 implement the ISO 14882 Standard C++ library as described in
40 chapters 17 through 27 and annex D. For those who want to see
41 exactly how far the project has come, or just want the latest
42 bleeding-edge code, the up-to-date source is available over
43 anonymous SVN, and can even be browsed over
44 the <ulink url="http://gcc.gnu.org/svn.html">web</ulink>.
49 <qandaentry id="faq.why">
52 Why should I use libstdc++?
57 The completion of the ISO C++ standardization gave the C++
58 community a powerful set of reuseable tools in the form of the C++
59 Standard Library. However, all existing C++ implementations are
60 (as the Draft Standard used to say) <quote>incomplet and
61 incorrekt</quote>, and many suffer from limitations of the compilers
65 The GNU compiler collection
66 (<command>gcc</command>, <command>g++</command>, etc) is widely
67 considered to be one of the leading compilers in the world. Its
68 development is overseen by the
69 <ulink url="http://gcc.gnu.org/">GCC team</ulink>. All of
70 the rapid development and near-legendary
71 <ulink url="http://gcc.gnu.org/buildstat.html">portability</ulink>
72 that are the hallmarks of an open-source project are being
76 That means that all of the Standard classes and functions will be
77 freely available and fully compliant. (Such as
78 <classname>string</classname>,
79 <classname>vector<></classname>, iostreams, and algorithms.)
80 Programmers will no longer need to <quote>roll their own</quote>
81 nor be worried about platform-specific incompatibilities.
86 <qandaentry id="faq.who">
89 Who's in charge of it?
94 The libstdc++ project is contributed to by several developers
95 all over the world, in the same way as GCC or Linux.
96 Benjamin Kosnik, Gabriel Dos Reis, Phil Edwards, Ulrich Drepper,
97 Loren James Rittle, and Paolo Carlini are the lead maintainers of
101 Development and discussion is held on the libstdc++ mailing
102 list. Subscribing to the list, or searching the list
103 archives, is open to everyone. You can read instructions for
104 doing so on the <ulink url="http://gcc.gnu.org/libstdc++/">homepage</ulink>.
105 If you have questions, ideas, code, or are just curious, sign up!
110 <qandaentry id="faq.when">
111 <question id="q-when">
113 When is libstdc++ going to be finished?
118 Nathan Myers gave the best of all possible answers, responding to
119 a Usenet article asking this question: <emphasis>Sooner, if you
125 <qandaentry id="faq.how">
126 <question id="q-how">
128 How do I contribute to the effort?
133 Here is <link linkend="appendix.contrib">a page devoted to
134 this topic</link>. Subscribing to the mailing list (see above, or
135 the homepage) is a very good idea if you have something to
136 contribute, or if you have spare time and want to
137 help. Contributions don't have to be in the form of source code;
138 anybody who is willing to help write documentation, for example,
139 or has found a bug in code that we all thought was working and is
140 willing to provide details, is more than welcome!
145 <qandaentry id="faq.whereis_old">
146 <question id="q-whereis_old">
148 What happened to the older libg++? I need that!
151 <answer id="a-whereis_old">
153 The most recent libg++ README states that libg++ is no longer
154 being actively maintained. It should not be used for new
155 projects, and is only being kicked along to support older code.
158 More information in the <link linkend="manual.appendix.porting.backwards">backwards compatibility documentation</link>
163 <qandaentry id="faq.more_questions">
164 <question id="q-more_questions">
166 What if I have more questions?
169 <answer id="a-more_questions">
171 If you have read the README file, and your question remains
172 unanswered, then just ask the mailing list. At present, you do not
173 need to be subscribed to the list to send a message to it. More
174 information is available on the homepage (including how to browse
175 the list archives); to send a message to the list,
176 use <email>libstdc++@gcc.gnu.org</email>.
180 If you have a question that you think should be included
181 here, or if you have a question <emphasis>about</emphasis> a question/answer
182 here, please send email to the libstdc++ mailing list, as above.
190 <qandadiv id="faq.license" xreflabel="License QA">
191 <title>License</title>
193 <qandaentry id="faq.license.what">
194 <question id="q-license.what">
196 What are the license terms for libstdc++?
199 <answer id="a-license.what">
201 See <link linkend="manual.intro.status.license">our license description</link>
202 for these and related questions.
207 <qandaentry id="faq.license.any_program">
208 <question id="q-license.any_program">
210 So any program which uses libstdc++ falls under the GPL?
213 <answer id="a-license.any_program">
215 No. The special exception permits use of the library in
216 proprietary applications.
222 <qandaentry id="faq.license.lgpl">
223 <question id="q-license.lgpl">
225 How is that different from the GNU {Lesser,Library} GPL?
228 <answer id="a-license.lgpl">
230 The LGPL requires that users be able to replace the LGPL code with a
231 modified version; this is trivial if the library in question is a C
232 shared library. But there's no way to make that work with C++, where
233 much of the library consists of inline functions and templates, which
234 are expanded inside the code that uses the library. So to allow people
235 to replace the library code, someone using the library would have to
236 distribute their own source, rendering the LGPL equivalent to the GPL.
241 <qandaentry id="faq.license.what_restrictions">
242 <question id="q-license.what_restrictions">
244 I see. So, what restrictions are there on programs that use the library?
247 <answer id="a-license.what_restrictions">
249 None. We encourage such programs to be released as open source,
250 but we won't punish you or sue you if you choose otherwise.
257 <!-- Installation -->
258 <qandadiv id="faq.installation" xreflabel="Installation">
259 <title>Installation</title>
261 <qandaentry id="faq.how_to_install">
262 <question id="q-how_to_install">
263 <para>How do I install libstdc++?
266 <answer id="a-how_to_install">
268 Often libstdc++ comes pre-installed as an integral part of many
269 existing Linux and Unix systems, as well as many embedded
270 development tools. It may be necessary to install extra
271 development packages to get the headers, or the documentation, or
272 the source: please consult your vendor for details.
275 To build and install from the GNU GCC sources, please consult the
276 <link linkend="manual.intro.setup">setup
277 documentation</link> for detailed
278 instructions. You may wish to browse those files ahead
279 of time to get a feel for what's required.
284 <qandaentry id="faq.how_to_get_sources">
285 <question id="q-how_to_get_sources">
286 <para>How does one get current libstdc++ sources?
289 <answer id="a-how_to_get_sources">
291 Libstdc++ sources for all official releases can be obtained as
292 part of the GCC sources, available from various sites and
293 mirrors. A full <ulink url="http://gcc.gnu.org/mirrors.html">list of
294 download sites</ulink> is provided on the main GCC site.
297 Current libstdc++ sources can always be checked out of the main
298 GCC source repository using the appropriate version control
299 tool. At this time, that tool
300 is <application>Subversion</application>.
303 <application>Subversion</application>, or <acronym>SVN</acronym>, is
304 one of several revision control packages. It was selected for GNU
305 projects because it's free (speech), free (beer), and very high
306 quality. The <ulink url="http://subversion.tigris.org"> Subversion
307 home page</ulink> has a better description.
310 The <quote>anonymous client checkout</quote> feature of SVN is
311 similar to anonymous FTP in that it allows anyone to retrieve
312 the latest libstdc++ sources.
316 see <ulink url="http://gcc.gnu.org/svn.html"><acronym>SVN</acronym>
322 <qandaentry id="faq.how_to_test">
323 <question id="q-how_to_test">
324 <para>How do I know if it works?
327 <answer id="a-how_to_test">
329 Libstdc++ comes with its own validation testsuite, which includes
330 conformance testing, regression testing, ABI testing, and
331 performance testing. Please consult the
332 <ulink url="http://gcc.gnu.org/install/test.html">testing
333 documentation</ulink> for more details.
336 If you find bugs in the testsuite programs themselves, or if you
337 think of a new test program that should be added to the suite,
338 <emphasis>please</emphasis> write up your idea and send it to the list!
343 <qandaentry id="faq.how_to_set_paths">
344 <question id="q-how_to_set_paths">
345 <para>How do I insure that the dynamically linked library will be found?
348 <answer id="a-how_to_set_paths">
350 Depending on your platform and library version, the error message might
351 be similar to one of the following:
355 ./a.out: error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory
357 /usr/libexec/ld-elf.so.1: Shared object "libstdc++.so.6" not found
361 This doesn't mean that the shared library isn't installed, only
362 that the dynamic linker can't find it. When a dynamically-linked
363 executable is run the linker finds and loads the required shared
364 libraries by searching a pre-configured list of directories. If
365 the directory where you've installed libstdc++ is not in this list
366 then the libraries won't be found. The simplest way to fix this is
367 to use the <literal>LD_LIBRARY_PATH</literal> environment variable,
368 which is a colon-separated list of directories in which the linker
369 will search for shared libraries:
373 LD_LIBRARY_PATH=${prefix}/lib:$LD_LIBRARY_PATH
374 export LD_LIBRARY_PATH
378 The exact environment variable to use will depend on your
379 platform, e.g. DYLD_LIBRARY_PATH for Darwin,
380 LD_LIBRARY_PATH_32/LD_LIBRARY_PATH_64 for Solaris 32-/64-bit,
381 LD_LIBRARYN32_PATH/LD_LIBRARY64_PATH for Irix N32/64-bit ABIs and
382 SHLIB_PATH for HP-UX.
385 See the man pages for <command>ld</command>, <command>ldd</command>
386 and <command>ldconfig</command> for more information. The dynamic
387 linker has different names on different platforms but the man page
388 is usually called something such as <filename>ld.so/rtld/dld.so</filename>.
391 Using LD_LIBRARY_PATH is not always the best solution, <link
392 linkend="manual.intro.using.linkage.dynamic">Finding Dynamic or Shared
393 Libraries</link> in the manual gives some alternatives.
398 <qandaentry id="faq.what_is_libsupcxx">
399 <question id="q-what_is_libsupcxx">
404 <answer id="a-what_is_libsupcxx">
406 If the only functions from <filename>libstdc++.a</filename>
407 which you need are language support functions (those listed in
408 <link linkend="std.support">clause 18</link> of the
409 standard, e.g., <function>new</function> and
410 <function>delete</function>), then try linking against
411 <filename>libsupc++.a</filename>, which is a subset of
412 <filename>libstdc++.a</filename>. (Using <command>gcc</command>
413 instead of <command>g++</command> and explicitly linking in
414 <filename>libsupc++.a</filename> via <literal>-lsupc++</literal>
415 for the final link step will do it). This library contains only
416 those support routines, one per object file. But if you are
417 using anything from the rest of the library, such as IOStreams
418 or vectors, then you'll still need pieces from
419 <filename>libstdc++.a</filename>.
424 <qandaentry id="faq.size">
425 <question id="q-size">
427 This library is HUGE!
432 Usually the size of libraries on disk isn't noticeable. When a
433 link editor (or simply <quote>linker</quote>) pulls things from a
434 static archive library, only the necessary object files are copied
435 into your executable, not the entire library. Unfortunately, even
436 if you only need a single function or variable from an object file,
437 the entire object file is extracted. (There's nothing unique to C++
438 or libstdc++ about this; it's just common behavior, given here
439 for background reasons.)
442 Some of the object files which make up libstdc++.a are rather large.
443 If you create a statically-linked executable with
444 <literal>-static</literal>, those large object files are suddenly part
445 of your executable. Historically the best way around this was to
446 only place a very few functions (often only a single one) in each
447 source/object file; then extracting a single function is the same
448 as extracting a single .o file. For libstdc++ this is only
449 possible to a certain extent; the object files in question contain
450 template classes and template functions, pre-instantiated, and
451 splitting those up causes severe maintenance headaches.
454 On supported platforms, libstdc++ takes advantage of garbage
455 collection in the GNU linker to get a result similar to separating
456 each symbol into a separate source and object files. On these platforms,
457 GNU ld can place each function and variable into its own
458 section in a .o file. The GNU linker can then perform garbage
459 collection on unused sections; this reduces the situation to only
460 copying needed functions into the executable, as before, but all
461 happens automatically.
469 <!-- Platform-Specific Issues -->
470 <qandadiv id="faq.platform-specific" xreflabel="Platform-Specific Issues">
471 <title>Platform-Specific Issues</title>
473 <qandaentry id="faq.other_compilers">
474 <question id="q-other_compilers">
476 Can libstdc++ be used with non-GNU compilers?
479 <answer id="a-other_compilers">
484 Since the goal of ISO Standardization is for all C++
485 implementations to be able to share code, libstdc++ should be
486 usable under any ISO-compliant compiler, at least in theory.
489 However, the reality is that libstdc++ is targeted and optimized
490 for GCC/g++. This means that often libstdc++ uses specific,
491 non-standard features of g++ that are not present in older
492 versions of proprietary compilers. It may take as much as a year or two
493 after an official release of GCC that contains these features for
494 proprietary tools support these constructs.
497 In the near past, specific released versions of libstdc++ have
498 been known to work with versions of the EDG C++ compiler, and
499 vendor-specific proprietary C++ compilers such as the Intel ICC
506 <qandaentry id="faq.solaris_long_long">
507 <question id="q-solaris_long_long">
509 No 'long long' type on Solaris?
512 <answer id="a-solaris_long_long">
514 By default we try to support the C99 <type>long long</type> type.
515 This requires that certain functions from your C library be present.
518 Up through release 3.0.2 the platform-specific tests performed by
519 libstdc++ were too general, resulting in a conservative approach
520 to enabling the <type>long long</type> code paths. The most
521 commonly reported platform affected was Solaris.
524 This has been fixed for libstdc++ releases greater than 3.0.3.
529 <qandaentry id="faq.predefined">
530 <question id="q-predefined">
532 <constant>_XOPEN_SOURCE</constant> and <constant>_GNU_SOURCE</constant> are always defined?
535 <answer id="a-predefined">
536 <para>On Solaris, g++ (but not gcc) always defines the preprocessor
537 macro <constant>_XOPEN_SOURCE</constant>. On GNU/Linux, the same happens
538 with <constant>_GNU_SOURCE</constant>. (This is not an exhaustive list;
539 other macros and other platforms are also affected.)
541 <para>These macros are typically used in C library headers, guarding new
542 versions of functions from their older versions. The C++ standard
543 library includes the C standard library, but it requires the C90
544 version, which for backwards-compatibility reasons is often not the
545 default for many vendors.
547 <para>More to the point, the C++ standard requires behavior which is only
548 available on certain platforms after certain symbols are defined.
549 Usually the issue involves I/O-related typedefs. In order to
550 ensure correctness, the compiler simply predefines those symbols.
552 <para>Note that it's not enough to #define them only when the library is
553 being built (during installation). Since we don't have an 'export'
554 keyword, much of the library exists as headers, which means that
555 the symbols must also be defined as your programs are parsed and
558 <para>To see which symbols are defined, look for CPLUSPLUS_CPP_SPEC in
559 the gcc config headers for your target (and try changing them to
560 see what happens when building complicated code). You can also run
561 <command>g++ -E -dM - < /dev/null"</command> to display
562 a list of predefined macros for any particular installation.
564 <para>This has been discussed on the mailing lists
565 <ulink url="http://gcc.gnu.org/cgi-bin/htsearch?method=and&format=builtin-long&sort=score&words=_XOPEN_SOURCE+Solaris">quite a bit</ulink>.
567 <para>This method is something of a wart. We'd like to find a cleaner
568 solution, but nobody yet has contributed the time.
574 <qandaentry id="faq.darwin_ctype">
575 <question id="q-darwin_ctype">
577 Mac OS X <filename class="headerfile">ctype.h</filename> is broken! How can I fix it?
580 <answer id="a-darwin_ctype">
581 <para>This is a long-standing bug in the OS X support. Fortunately,
582 the patch is quite simple, and well-known.
583 <ulink url="http://gcc.gnu.org/ml/gcc/2002-03/msg00817.html"> Here's a
584 link to the solution</ulink>.
590 <qandaentry id="faq.threads_i386">
591 <question id="q-threads_i386">
593 Threading is broken on i386?
596 <answer id="a-threads_i386">
599 <para>Support for atomic integer operations is/was broken on i386
600 platforms. The assembly code accidentally used opcodes that are
601 only available on the i486 and later. So if you configured GCC
602 to target, for example, i386-linux, but actually used the programs
603 on an i686, then you would encounter no problems. Only when
604 actually running the code on a i386 will the problem appear.
606 <para>This is fixed in 3.2.2.
612 <qandaentry id="faq.atomic_mips">
613 <question id="q-atomic_mips">
615 MIPS atomic operations
618 <answer id="a-atomic_mips">
620 The atomic locking routines for MIPS targets requires MIPS II
621 and later. A patch went in just after the 3.3 release to
622 make mips* use the generic implementation instead. You can also
623 configure for mipsel-elf as a workaround.
626 The mips*-*-linux* port continues to use the MIPS II routines, and more
627 work in this area is expected.
632 <qandaentry id="faq.linux_glibc">
633 <question id="q-linux_glibc">
635 Recent GNU/Linux glibc required?
638 <answer id="a-linux_glibc">
639 <para>When running on GNU/Linux, libstdc++ 3.2.1 (shared library version
640 5.0.1) and later uses localization and formatting code from the system
641 C library (glibc) version 2.2.5 which contains necessary bugfixes.
642 Most GNU/Linux distros make more recent versions available now.
644 <para>The guideline is simple: the more recent the C++ library, the
645 more recent the C library. (This is also documented in the main
646 GCC installation instructions.)
652 <qandaentry id="faq.freebsd_wchar">
653 <question id="q-freebsd_wchar">
655 Can't use wchar_t/wstring on FreeBSD
658 <answer id="a-freebsd_wchar">
660 Older versions of FreeBSD's C library do not have sufficient
661 support for wide character functions, and as a result the
662 libstdc++ configury decides that wchar_t support should be
663 disabled. In addition, the libstdc++ platform checks that
664 enabled <type>wchar_t</type> were quite strict, and not granular
665 enough to detect when the minimal support to
666 enable <type>wchar_t</type> and C++ library structures
667 like <classname>wstring</classname> were present. This impacted Solaris,
668 Darwin, and BSD variants, and is fixed in libstdc++ versions post 4.1.0.
679 <qandadiv id="faq.known_bugs" xreflabel="Known Bugs">
680 <title>Known Bugs</title>
682 <qandaentry id="faq.what_works">
683 <question id="q-what_works">
688 <answer id="a-what_works">
690 Short answer: Pretty much everything <emphasis>works</emphasis>
691 except for some corner cases. Support for localization
692 in <classname>locale</classname> may be incomplete on non-GNU
693 platforms. Also dependant on the underlying platform is support
694 for <type>wchar_t</type> and <type>long
695 long</type> specializations, and details of thread support.
698 Long answer: See the implementation status pages for
699 <link linkend="status.iso.1998">C++98</link>,
700 <link linkend="status.iso.tr1">TR1</link>, and
701 <link linkend="status.iso.200x">C++0x</link>.
706 <qandaentry id="faq.standard_bugs">
707 <question id="q-standard_bugs">
709 Bugs in the ISO C++ language or library specification
712 <answer id="a-standard_bugs">
714 Unfortunately, there are some.
717 For those people who are not part of the ISO Library Group
718 (i.e., nearly all of us needing to read this page in the first
719 place), a public list of the library defects is occasionally
720 published <ulink url="http://www.open-std.org/jtc1/sc22/wg21/">here</ulink>.
721 Some of these issues have resulted in code changes in libstdc++.
724 If you think you've discovered a new bug that is not listed,
725 please post a message describing your problem
726 to <email>libstdc++@gcc.gnu.org</email> or the Usenet group
727 comp.lang.c++.moderated.
732 <qandaentry id="faq.compiler_bugs">
733 <question id="q-compiler_bugs">
735 Bugs in the compiler (gcc/g++) and not libstdc++
738 <answer id="a-compiler_bugs">
740 On occasion, the compiler is wrong. Please be advised that this
741 happens much less often than one would think, and avoid jumping to
745 First, examine the ISO C++ standard. Second, try another compiler
746 or an older version of the GNU compilers. Third, you can find more
747 information on the libstdc++ and the GCC mailing lists: search
748 these lists with terms describing your issue.
751 Before reporting a bug, please examine the
752 <ulink url="http://gcc.gnu.org/bugs.html">bugs database</ulink> with the
753 category set to <quote>g++</quote>.
760 <!-- Known Non-Bugs -->
761 <qandadiv id="faq.known_non-bugs" xreflabel="Known Non-Bugs">
762 <title>Known Non-Bugs</title>
764 <qandaentry id="faq.stream_reopening_fails">
765 <question id="q-stream_reopening_fails">
767 Reopening a stream fails
770 <answer id="a-stream_reopening_fails">
772 One of the most-reported non-bug reports. Executing a sequence like:
776 #include <fstream>
778 std::fstream fs(<quote>a_file</quote>);
780 // . do things with fs...
783 fs.open(<quote>a_new_file</quote>);
787 All operations on the re-opened <varname>fs</varname> will fail, or at
788 least act very strangely. Yes, they often will, especially if
789 <varname>fs</varname> reached the EOF state on the previous file. The
790 reason is that the state flags are <emphasis>not</emphasis> cleared
791 on a successful call to open(). The standard unfortunately did
792 not specify behavior in this case, and to everybody's great sorrow,
793 the <link linkend="manual.intro.status.bugs">proposed LWG resolution in
794 DR #22</link> is to leave the flags unchanged. You must insert a call
795 to <function>fs.clear()</function> between the calls to close() and open(),
796 and then everything will work like we all expect it to work.
797 <emphasis>Update:</emphasis> for GCC 4.0 we implemented the resolution
798 of <link linkend="manual.intro.status.bugs">DR #409</link> and open()
799 now calls <function>clear()</function> on success!
804 <qandaentry id="faq.wefcxx_verbose">
805 <question id="q-wefcxx_verbose">
807 -Weffc++ complains too much
810 <answer id="a-wefcxx_verbose">
812 Many warnings are emitted when <literal>-Weffc++</literal> is used. Making
813 libstdc++ <literal>-Weffc++</literal>-clean is not a goal of the project,
814 for a few reasons. Mainly, that option tries to enforce
815 object-oriented programming, while the Standard Library isn't
816 necessarily trying to be OO.
819 We do, however, try to have libstdc++ sources as clean as possible. If
820 you see some simple changes that pacify <literal>-Weffc++</literal>
821 without other drawbacks, send us a patch.
826 <qandaentry id="faq.ambiguous_overloads">
827 <question id="q-ambiguous_overloads">
829 Ambiguous overloads after including an old-style header
832 <answer id="a-ambiguous_overloads">
834 Another problem is the <literal>rel_ops</literal> namespace and the template
835 comparison operator functions contained therein. If they become
836 visible in the same namespace as other comparison functions
837 (e.g., <quote>using</quote> them and the <iterator> header),
838 then you will suddenly be faced with huge numbers of ambiguity
839 errors. This was discussed on the -v3 list; Nathan Myers
840 <ulink url="http://gcc.gnu.org/ml/libstdc++/2001-01/msg00247.html">sums
841 things up here</ulink>. The collisions with vector/string iterator
842 types have been fixed for 3.1.
847 <qandaentry id="faq.v2_headers">
848 <question id="q-v2_headers">
850 The g++-3 headers are <emphasis>not ours</emphasis>
853 <answer id="a-v2_headers">
855 If you have found an extremely broken header file which is
856 causing problems for you, look carefully before submitting a
857 "high" priority bug report (which you probably
858 shouldn't do anyhow; see the last paragraph of the page
859 describing <ulink url="http://gcc.gnu.org/bugs.html">the GCC
860 bug database</ulink>).
863 If the headers are in <filename>${prefix}/include/g++-3</filename>, or
864 if the installed library's name looks like
865 <filename>libstdc++-2.10.a</filename> or
866 <filename>libstdc++-libc6-2.10.so</filename>, then you are using the
867 old libstdc++-v2 library, which is nonstandard and
868 unmaintained. Do not report problems with -v2 to the -v3
872 For GCC versions 3.0 and 3.1 the libstdc++ header files are
873 installed in <filename>${prefix}/include/g++-v3</filename> (see the
874 'v'?). Starting with version 3.2 the headers are installed in
875 <filename>${prefix}/include/c++/${version}</filename> as this prevents
876 headers from previous versions being found by mistake.
882 <qandaentry id="faq.boost_concept_checks">
883 <question id="q-boost_concept_checks">
885 Errors about <emphasis>*Concept</emphasis> and
886 <emphasis>constraints</emphasis> in the STL
889 <answer id="a-boost_concept_checks">
891 If you see compilation errors containing messages about
892 <errortext>foo Concept </errortext>and something to do with a
893 <errortext>constraints</errortext> member function, then most
894 likely you have violated one of the requirements for types used
895 during instantiation of template containers and functions. For
896 example, EqualityComparableConcept appears if your types must be
897 comparable with == and you have not provided this capability (a
898 typo, or wrong visibility, or you just plain forgot, etc).
901 More information, including how to optionally enable/disable the
902 checks, is available in the
903 <link linkend="std.diagnostics.concept_checking">Diagnostics</link>.
904 chapter of the manual.
909 <qandaentry id="faq.dlopen_crash">
910 <question id="q-dlopen_crash">
912 Program crashes when using library code in a
913 dynamically-loaded library
916 <answer id="a-dlopen_crash">
918 If you are using the C++ library across dynamically-loaded
919 objects, make certain that you are passing the correct options
920 when compiling and linking:
924 // compile your library components
930 // create your library
931 g++ -fPIC -shared -rdynamic -o libfoo.so a.o b.o ... z.o
933 // link the executable
934 g++ -fPIC -rdynamic -o foo ... -L. -lfoo -ldl
939 <qandaentry id="faq.memory_leaks">
940 <question id="q-memory_leaks">
942 <quote>Memory leaks</quote> in containers
945 <answer id="a-memory_leaks">
947 A few people have reported that the standard containers appear
948 to leak memory when tested with memory checkers such as
949 <ulink url="http://valgrind.org/">valgrind</ulink>.
950 The library's default allocators keep free memory in a pool
951 for later reuse, rather than returning it to the OS. Although
952 this memory is always reachable by the library and is never
953 lost, memory debugging tools can report it as a leak. If you
954 want to test the library for memory leaks please read
955 <link linkend="debug.memory">Tips for memory leak hunting</link>
961 <qandaentry id="faq.list_size_on">
962 <question id="q-list_size_on">
964 list::size() is O(n)!
967 <answer id="a-list_size_on">
970 the <link linkend="std.containers">Containers</link>
976 <qandaentry id="faq.easy_to_fix">
977 <question id="q-easy_to_fix">
979 Aw, that's easy to fix!
982 <answer id="a-easy_to_fix">
984 If you have found a bug in the library and you think you have
985 a working fix, then send it in! The main GCC site has a page
986 on <ulink url="http://gcc.gnu.org/contribute.html">submitting
987 patches</ulink> that covers the procedure, but for libstdc++ you
988 should also send the patch to our mailing list in addition to
989 the GCC patches mailing list. The libstdc++
990 <link linkend="appendix.contrib">contributors' page</link>
991 also talks about how to submit patches.
994 In addition to the description, the patch, and the ChangeLog
995 entry, it is a Good Thing if you can additionally create a small
996 test program to test for the presence of the bug that your
997 patch fixes. Bugs have a way of being reintroduced; if an old
998 bug creeps back in, it will be caught immediately by the
999 <ulink url="#2_4">testsuite</ulink> -- but only if such a test exists.
1007 <!-- Miscellaneous -->
1008 <qandadiv id="faq.misc" xreflabel="Miscellaneous">
1009 <title>Miscellaneous</title>
1011 <qandaentry id="faq.iterator_as_pod">
1012 <question id="faq.iterator_as_pod_q">
1014 string::iterator is not char*; vector<T>::iterator is not T*
1017 <answer id="faq.iterator_as_pod_a">
1019 If you have code that depends on container<T> iterators
1020 being implemented as pointer-to-T, your code is broken. It's
1021 considered a feature, not a bug, that libstdc++ points this out.
1024 While there are arguments for iterators to be implemented in
1025 that manner, A) they aren't very good ones in the long term,
1026 and B) they were never guaranteed by the Standard anyway. The
1027 type-safety achieved by making iterators a real class rather
1028 than a typedef for <type>T*</type> outweighs nearly all opposing
1032 Code which does assume that a vector iterator <varname>i</varname>
1033 is a pointer can often be fixed by changing <varname>i</varname> in
1034 certain expressions to <varname>&*i</varname>. Future revisions
1035 of the Standard are expected to bless this usage for
1036 vector<> (but not for basic_string<>).
1041 <qandaentry id="faq.what_is_next">
1042 <question id="q-what_is_next">
1044 What's next after libstdc++?
1047 <answer id="a-what_is_next">
1049 Hopefully, not much. The goal of libstdc++ is to produce a
1050 fully-compliant, fully-portable Standard Library. After that,
1051 we're mostly done: there won't <emphasis>be</emphasis> any
1052 more compliance work to do.
1055 There is an effort underway to add significant extensions to
1056 the standard library specification. The latest version of
1057 this effort is described in
1058 <ulink url="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1836.pdf">
1059 The C++ Library Technical Report 1</ulink>.
1064 <qandaentry id="faq.sgi_stl">
1065 <question id="q-sgi_stl">
1067 What about the STL from SGI?
1070 <answer id="a-sgi_stl">
1072 The <ulink url="http://www.sgi.com/tech/stl/">STL from SGI</ulink>,
1073 version 3.3, was the final merge of the STL codebase. The
1074 code in libstdc++ contains many fixes and changes, and
1075 the SGI code is no longer under active
1076 development. We expect that no future merges will take place.
1079 In particular, <classname>string</classname> is not from SGI and makes no
1080 use of their "rope" class (which is included as an
1081 optional extension), nor is <classname>valarray</classname> and some others.
1082 Classes like <classname>vector<></classname> are, but have been
1083 extensively modified.
1086 More information on the evolution of libstdc++ can be found at the
1087 <link linkend="appendix.porting.api">API
1089 and <link linkend="manual.appendix.porting.backwards">backwards
1090 compatibility</link> documentation.
1093 The FAQ for SGI's STL (one jump off of their main page) is
1094 still recommended reading.
1099 <qandaentry id="faq.extensions_and_backwards_compat">
1100 <question id="q-extensions_and_backwards_compat">
1102 Extensions and Backward Compatibility
1105 <answer id="a-extensions_and_backwards_compat">
1107 See the <link linkend="manual.appendix.porting.backwards">link</link> on backwards compatibility and <link linkend="appendix.porting.api">link</link> on evolution.
1112 <qandaentry id="faq.tr1_support">
1113 <question id="q-tr1_support">
1115 Does libstdc++ support TR1?
1118 <answer id="a-tr1_support">
1123 The C++ Standard Library Technical Report adds many new features to
1124 the library. The latest version of this effort is described in
1126 "http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1836.pdf">
1127 Technical Report 1</ulink>.
1130 The implementation status of TR1 in libstdc++ can be tracked <link
1131 linkend="status.iso.tr1">on the TR1 status
1137 <qandaentry id="faq.get_iso_cxx">
1138 <question id="q-get_iso_cxx">
1139 <para>How do I get a copy of the ISO C++ Standard?
1142 <answer id="a-get_iso_cxx">
1144 Copies of the full ISO 14882 standard are available on line via
1145 the ISO mirror site for committee members. Non-members, or those
1146 who have not paid for the privilege of sitting on the committee
1147 and sustained their two-meeting commitment for voting rights, may
1148 get a copy of the standard from their respective national
1149 standards organization. In the USA, this national standards
1150 organization is ANSI and their website is
1151 right <ulink url="http://www.ansi.org">here</ulink>. (And if
1152 you've already registered with them, clicking this link will take
1153 you to directly to the place where you can
1154 <ulink url="http://webstore.ansi.org/RecordDetail.aspx?sku=ISO%2FIEC+14882:2003">buy the standard on-line</ulink>.
1157 Who is your country's member body? Visit the
1158 <ulink url="http://www.iso.ch/">ISO homepage</ulink> and find out!
1161 The 2003 version of the standard (the 1998 version plus TC1) is
1162 available in print, ISBN 0-470-84674-7.
1167 <qandaentry id="faq.what_is_abi">
1168 <question id="q-what_is_abi">
1170 What's an ABI and why is it so messy?
1173 <answer id="a-what_is_abi">
1175 <acronym>ABI</acronym> stands for <quote>Application Binary
1176 Interface</quote>. Conventionally, it refers to a great
1177 mass of details about how arguments are arranged on the call
1178 stack and/or in registers, and how various types are arranged
1179 and padded in structs. A single CPU design may suffer
1180 multiple ABIs designed by different development tool vendors
1181 who made different choices, or even by the same vendor for
1182 different target applications or compiler versions. In ideal
1183 circumstances the CPU designer presents one ABI and all the
1184 OSes and compilers use it. In practice every ABI omits
1185 details that compiler implementers (consciously or
1186 accidentally) must choose for themselves.
1189 That ABI definition suffices for compilers to generate code so a
1190 program can interact safely with an OS and its lowest-level libraries.
1191 Users usually want an ABI to encompass more detail, allowing libraries
1192 built with different compilers (or different releases of the same
1193 compiler!) to be linked together. For C++, this includes many more
1194 details than for C, and CPU designers (for good reasons elaborated
1195 below) have not stepped up to publish C++ ABIs. The details include
1196 virtual function implementation, struct inheritance layout, name
1197 mangling, and exception handling. Such an ABI has been defined for
1198 GNU C++, and is immediately useful for embedded work relying only on
1199 a <quote>free-standing implementation</quote> that doesn't include (much
1200 of) the standard library. It is a good basis for the work to come.
1203 A useful C++ ABI must also incorporate many details of the standard
1204 library implementation. For a C ABI, the layouts of a few structs
1205 (such as FILE, stat, jmpbuf, and the like) and a few macros suffice.
1206 For C++, the details include the complete set of names of functions
1207 and types used, the offsets of class members and virtual functions,
1208 and the actual definitions of all inlines. C++ exposes many more
1209 library details to the caller than C does. It makes defining
1210 a complete ABI a much bigger undertaking, and requires not just
1211 documenting library implementation details, but carefully designing
1212 those details so that future bug fixes and optimizations don't
1213 force breaking the ABI.
1216 There are ways to help isolate library implementation details from the
1217 ABI, but they trade off against speed. Library details used in
1218 inner loops (e.g., getchar) must be exposed and frozen for all
1219 time, but many others may reasonably be kept hidden from user code,
1220 so they may later be changed. Deciding which, and implementing
1221 the decisions, must happen before you can reasonably document a
1222 candidate C++ ABI that encompasses the standard library.
1227 <qandaentry id="faq.size_equals_capacity">
1228 <question id="q-size_equals_capacity">
1230 How do I make std::vector<T>::capacity() == std::vector<T>::size?
1233 <answer id="a-size_equals_capacity">
1235 The standard idiom for deallocating a <classname>vector<T></classname>'s
1236 unused memory is to create a temporary copy of the vector and swap their
1237 contents, e.g. for <classname>vector<T> v</classname>
1240 std::vector<T>(v).swap(v);
1243 The copy will take O(n) time and the swap is constant time.
1246 See <link linkend="strings.string.shrink">Shrink-to-fit
1247 strings</link> for a similar solution for strings.
1255 <!-- FAQ ends here -->